Update to 0.4.0-dev spec #1

emery commented 2022年04月28日 00:18:53 +02:00

No description provided.

emery commented 2022年04月28日 18:01:56 +02:00

I just heard back from the hashfs users that they are migrating away from hashfs. :|

I just heard back from the hashfs users that they are migrating away from hashfs. :|

pukkamustard commented 2022年05月05日 12:57:53 +02:00

Oops, sorry for closing. Clicked the wrong button by mistake.

Oops, sorry for closing. Clicked the wrong button by mistake.

pukkamustard commented 2022年05月05日 13:09:32 +02:00

Some more changes coming to 0.4.0 so I suggest keeping this open until 0.4.0 drops.

Some more changes coming to 0.4.0 so I suggest keeping this open until 0.4.0 drops.

pukkamustard commented 2022年05月10日 20:33:02 +02:00

Just saw that python-eris now has two implementations of the encoding! :)

Just saw that python-eris now has two implementations of the encoding! :)

pukkamustard commented 2022年05月10日 20:41:18 +02:00

Pushed another update of the implementation at eris/encode.py: 484f0d55d7

Working on a decoder now.

Pushed another update of the implementation at eris/encode.py: https://codeberg.org/eris/python-eris/commit/484f0d55d7230a5c5ecd6df56dd84302d797a79d Working on a decoder now.

pukkamustard commented 2022年05月11日 10:53:21 +02:00

I've moved the hashlib interface to eris/encode.py: a5bd0d2a80, added one more change from erisx3 and merged to main.

I'm sorry to have dropped your very neat implementation in hashlib_extension.py for the more verbose one in eris/encode.py. For anybody interested, it's very concise (less than 100 lines of code) and re-uses the first part of the buffer for constructing internal nodes: https://codeberg.org/eris/python-eris/src/branch/erisx3/eris/hashlib_extension.py. I don't think the re-use of the buffer improves performance in Python, as Python bytes are immutable. But it's a neat trick that should be used in a C (and C-esque languages) implementation.

Since hashfs users are migrating away from hashfs I've also droped the hashfs tests.

I've moved the hashlib interface to eris/encode.py: https://codeberg.org/eris/python-eris/commit/a5bd0d2a80d483f85edf6544f8ba5c321e17dcfb, added one more change from erisx3 and merged to main. I'm sorry to have dropped your very neat implementation in `hashlib_extension.py` for the more verbose one in `eris/encode.py`. For anybody interested, it's very concise (less than 100 lines of code) and re-uses the first part of the buffer for constructing internal nodes: https://codeberg.org/eris/python-eris/src/branch/erisx3/eris/hashlib_extension.py. I don't think the re-use of the buffer improves performance in Python, as Python bytes are immutable. But it's a neat trick that should be used in a C (and C-esque languages) implementation. Since hashfs users are migrating away from hashfs I've also droped the hashfs tests.

emery commented 2022年05月11日 16:44:20 +02:00

Deduping the encoders is fine by me, that method came from the Squeak implementation and now it's the one I'm able to keep in my mind. I did have difficulty working with Python buffers so I can imagine it's not ideal.

If that is the most compact algorithm we can find then I can describe it in the spec, either as pseudocode or in an appendix.

Deduping the encoders is fine by me, that method came from the Squeak implementation and now it's the one I'm able to keep in my mind. I did have difficulty working with Python buffers so I can imagine it's not ideal. If that is the most compact algorithm we can find then I can describe it in the spec, either as pseudocode or in an appendix.